Anthropic’s Claude Opus 5.5, released in September 2026, established a strong baseline on HLE-Diamond, the 1,000-question refined Humanity’s Last Exam benchmark focused on expert-level reasoning and knowledge without tools. It reached 54.6% at standard high reasoning effort and up to 67.9% at maximum effort on Artificial Analysis and Scale leaderboards, trailing only certain Gemini 4 variants while outperforming most GPT-6 models on the knowledge partition. Traders see this performance, combined with Anthropic’s rapid Opus cadence (Opus 5 in July followed by 5.5 in September), as evidence that the next 5.6+ release will clear high thresholds when submitted. Ongoing competition from Google and OpenAI, plus Anthropic’s focus on agentic coding and test-time compute scaling, reinforces expectations for continued gains ahead of any potential early-2027 launch.